gpt2-finetuned-academic-topics

This model is a fine-tuned version of gpt2 on a dataset of sequences of science, technology, engineering and mathematics academic topics/tags which a user has used on their CiteULike or Google Scholar profiles.

Please contact brichards88@uri.edu for questions or inquiries.

It achieves the following results on the evaluation set:

• Train Loss: 3.3216
• Validation Loss: 3.2215
• Epoch: 4

Model description

Give a sequence of topics, i.e.: "machine learning, deep learning, chemistry, evolution" the model will continue the sequence, effectively recommending/generating new topics that might be of interest.

Intended uses & limitations

The model is not guaranteed to generate a real topic or even a real word/words as output.

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• optimizer: {'name': 'AdamWeightDecay', 'learning_rate': 2e-05, 'decay': 0.0, 'beta_1': 0.9, 'beta_2': 0.999, 'epsilon': 1e-07, 'amsgrad': False, 'weight_decay_rate': 0.01}
• training_precision: float32

Training results

Train Loss	Validation Loss	Epoch
4.7873	4.2950	0
4.1032	3.8203	1
3.7363	3.5614	2
3.4999	3.3740	3
3.3216	3.2215	4

Framework versions

• Transformers 4.18.0
• TensorFlow 2.8.0
• Datasets 2.1.0
• Tokenizers 0.12.1